Oncofusion

start / graph / gene-flt3

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q5c-interactome2

FLT3 / CD135 — the strongest non-tether co-dependency, and a target the field may have written off on a badly executed trial

also called FLT3 · CD135 · FLK2 · STK1 · fms-like tyrosine kinase 3

FLT3 is a growth-signal receptor on the surface of young blood cells, and KMT2A-fusion leukemias switch it up to very high levels. Cells that need menin also need FLT3 more than almost any other gene we tested, which sits awkwardly with a large infant trial of a FLT3 drug that failed.

The same thing, for a scientist

FLT3 (P36888, 993 aa) is a class III receptor tyrosine kinase restricted to haematopoietic progenitors. In DepMap it is an unusually selective dependency: depleted in fewer than 1% of 1,243 cell lines overall, yet the six most FLT3-depleted leukemia lines are all KMT2A-rearranged. Leukemia-restricted co-dependency with MEN1 is r = 0.645, the fourth strongest signal in this project's bootstrap and the strongest outside the menin-LEDGF-KMT2A-DOT1L tether itself. The apparent conflict with the negative COG AALL0631 lestaurtinib trial resolves once the trial's own pharmacodynamic data are read.

loading 6JQR…
PDB 6JQR
Unlike every other protein in this beat, FLT3 already has approved drugs that bind it, with the structures to prove it.
Tell me more about this picture

6JQR: FLT3 kinase domain with gilteritinib, 2.2 A X-ray.

Tell me more about this picture

Gilteritinib is a type I inhibitor and binds the active conformation, so it retains activity against activation-loop mutants that defeat the type II inhibitor quizartinib (4XUF).

Tell me more about this picture

In KMT2A-rearranged infant ALL the lesion is overexpression of the wild-type receptor rather than either mutation class, so allele coverage is not the differentiator; potency, paediatric exposure and a measured pharmacodynamic endpoint are.

What it looks like
FLT3993 aa
Ig-like C2-type (extracellular)Ig-like C2-type (extracellular) · residues 253–343Juxtamembrane autoinhibitory segment (UniProt-flagged residues) · residues 591–597Protein kinase domain (split)Protein kinase domain (split) · residues 610–9430200400600800
loading AF-P36888-F1…
AlphaFold AF-P36888-F1 ↗
colourshape

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
Does the FLT3 tension reconcile? Yes — and the trial itself supplies the reconciliation2 deeper

The tension this graph recorded was three-sided: FLT3 overexpression is near-universal in KMT2A-r, activating mutation is far rarer than the field assumes (about 3% in infants, no ITD observed), and a large randomised FLT3-inhibitor trial in exactly this population was negative. A strong co-dependency signal looked hard to square with that.

It squares. COG AALL0631 randomised 121 KMT2A-r infants and found no EFS difference: 36 ± 6% with lestaurtinib versus 39 ± 7% with chemotherapy alone, p = 0.67. But the trial measured whether the drug actually inhibited FLT3, and it mostly did not. Among lestaurtinib-treated patients, those whose plasma achieved FLT3 inhibition had 3-year EFS 59 ± 10% against 28 ± 7% for the uninhibited (p = 0.009). Those whose blasts were sensitive to FLT3 inhibition ex vivo had 52 ± 8% against 5 ± 5% for the resistant (p < 0.001). The 17 patients who were both inhibited and sensitive had EFS 88 ± 8% 2.

So AALL0631 is not evidence that FLT3 is not a dependency in KMT2A-r infant ALL. It is evidence that lestaurtinib, given at that schedule, failed to hit the target in most patients. Those are different conclusions, and only the second is supported.

Why this reading must be held with discipline

The pharmacodynamic and ex-vivo-sensitivity comparisons are post-hoc subgroup analyses within one randomised arm, and they are not themselves randomised. Two confounders are obvious and neither is excluded by the paper.

First, achieving plasma FLT3 inhibition depends on drug exposure, which depends on hepatic and renal function, nutrition and compliance — all of which are independently prognostic in a sick infant. The inhibited group may simply be the less ill group.

Second, ex vivo FLT3-inhibitor sensitivity is plausibly a biomarker of disease biology in its own right. A blast population that dies in a dish is a different disease from one that does not, irrespective of what killed it. The 5 ± 5% EFS in the ex-vivo-resistant group is a catastrophic prognosis that no drug assignment would plausibly produce on its own.

The honest statement is: the trial's failure is fully compatible with FLT3 being a real dependency, and the trial's subgroups are suggestive but cannot establish it. The experiment that would settle it is a potent, selective, modern FLT3 inhibitor — gilteritinib or quizartinib — dosed to a measured pharmacodynamic endpoint, with patients selected on FLT3 expression rather than mutation.

Correction, 2026-09-20: three of those five conditions were met in 2016, and the result is a negative. The sentence that stood here said flatly that the trial had not been run. PMID 26920889 ran quizartinib — the paper's own words are "a potent and selective class III receptor tyrosine kinase (RTK) FLT3 inhibitor" — in "children with relapsed AML or MLL-rearranged ALL", dosed to a measured pharmacodynamic endpoint: "Plasma inhibitory assay (PIA) testing was performed weekly to determine biologic activity", achieving "near complete inhibition of FLT3 phosphorylation in all patients".

What has still not been run is the expression-selected version, in infants, frontline. The 2016 study enrolled by diagnosis and reported its subgroups by mutation, in relapsed disease, on top of high-dose cytarabine and etoposide.

And the part that matters most is the part that argues against this dossier. The paper reports that "FLT3-ITD patients, but not FLT3 wild-type (WT) patients, had significantly lower blast counts post-quizartinib" — while FLT3 phosphorylation was completely inhibited in all patients. KMT2A-rearranged ALL is overwhelmingly FLT3 wild-type-overexpressing, not ITD. So the one published trial that gave a modern selective FLT3 inhibitor to this population, and proved on assay that it hit the target, found no blast response in the mutation class this population belongs to. That is the sharpest published counter-argument to the FLT3-expression hypothesis in this entry, and the entry did not have it.

The mutation paradox also resolves, and the resolution was published in 2003

The co-dependency is on FLT3 the protein, not on FLT3 the mutant allele. Armstrong and colleagues reported in the original paper that established FLT3 as a target in this disease that PKC412 was differentially cytotoxic to MLL-translocated cells in which FLT3 was activated by overexpression of the wild-type receptor or by mutation 3. High-level wild-type receptor expression produces ligand-independent phosphorylation. A cell can be addicted to a wild-type kinase it has too much of.

The DepMap data say the same thing directly. Of the six most FLT3-depleted leukemia lines, all six are KMT2A-rearranged, and three of them — SEM, SEMK2 and HB11;19 — carry no recorded FLT3 internal tandem duplication 14. The FLT3 requirement in KMT2A-rearranged cells is not confined to FLT3-mutant cells.

This also explains why the field's selection biomarker has been wrong. Selecting on FLT3 mutation would enrol roughly 3% of infant KMT2A-r patients; selecting on expression would enrol nearly all of them.

The dependency profile, which is the best in this whole beat

FLT3 is depleted in 0.9% of 1,243 DepMap lines — the most selective profile of any gene in this domain. Mean Chronos is -0.145 overall and -0.279 across 57 leukemia lines 1. Compare that with the elongation machinery this graph has already written off: BRD4 -1.01 and depleted in 94.7% of lines, PRMT1 -1.40 and 98.7%, CDK9 -1.54 and 99.3%.

The most depleted leukemia lines are a near-pure KMT2A-rearranged set 14:

LineFusionFLT3-ITD?Chronos
MOLM-13KMT2A::MLLT3yes-2.181
SEMK2KMT2A::AFF1none recorded-1.715
HB11;19KMT2A::MLLT1none recorded-1.617
MV4-11KMT2A::AFF1yes-1.509
SEMKMT2A::AFF1none recorded-0.982
MOLM-14KMT2A::MLLT3—-0.970

FLT3 correlates with MEN1 at r = 0.645 (95% CI 0.362–0.810) across 54 leukemia lines, surviving Holm correction, but at only r = 0.15 across the full 1,208-line panel 5. As with DPY30, the association is a leukemia-context phenomenon rather than a constitutive one.

Tolerability, and why this is the most favourable dossier in the beat1 deeper

Mouse. Flt3-null mice develop into healthy adults with normal mature haematopoietic populations. They have a specific deficiency in primitive B-lymphoid progenitors, and their stem cells reconstitute T-cell and myeloid lineages poorly on transplantation. Combined c-kit/flt3 deficiency is far more severe and postnatally lethal 6. A viable, healthy knockout is a strong tolerability signal, and it is better than anything in this beat except LEDGF.

Expression. FLT3 is tissue-enhanced and effectively confined to haematopoietic tissue: bone marrow 5.0, lymph node 6.1, spleen 5.3, thymus 4.3 nTPM, against heart muscle 0.1, liver 0.4 and testis 0.6 1. There is almost no non-haematopoietic tissue to damage.

Human constraint. pLI 0.998, LOEUF 0.465, 46 observed against 126.6 expected loss-of-function variants 1. This looks constrained, but FLT3 is a proto-oncogene whose germline gain-of-function alleles are selected against; as with MEN1, constraint and cell-essentiality are answering different questions here.

Chemical matter. Unlike everything else in this beat, FLT3 is thoroughly drugged. Gilteritinib and quizartinib are FDA-approved FLT3 inhibitors in adult AML, and co-crystal structures exist at 2.2 Å (6JQR, gilteritinib) and 3.2 Å (4XUF, quizartinib) 7. The molecules exist; the paediatric KMT2A-r trial designed around expression does not.

The domain map, and where the drugs bind

FLT3 is 993 residues: five extracellular Ig-like domains (UniProt annotates the C2-type domain at 253–343), a transmembrane segment, a juxtamembrane autoinhibitory region in which UniProt flags residues 591–597 as important for keeping the kinase off in the absence of ligand, and a split protein-kinase domain at 610–943 8.

The geometry matters for the paediatric case. Internal tandem duplications land in the juxtamembrane region and relieve autoinhibition; activation-loop mutations such as D835 and I836del sit inside the kinase domain. Quizartinib is a type II inhibitor that requires the inactive conformation and is therefore defeated by activation-loop mutations; gilteritinib is type I and is not. In a population where the lesion is overexpression of a wild-type receptor rather than either mutation class, both should work on the kinase, and the choice turns on potency and paediatric tolerability rather than on allele coverage.

The epigenetic dimension runs in both directions, which is why 'FLT3 is a receptor kinase' is the wrong answer

FLT3 has no chromatin-modifying domain, its substrates are non-histone, and FLT3-ITD[tiab] AND H3K9[tiab] returns 0 records. The tempting entry is therefore not applicable, as this entry already correctly records for the subnuclear question. That would be wrong, because FLT3 sits at both ends of an epigenetic axis in this disease.

Inbound: the FLT3 locus, and the best of it is human and B-lineage. Integrative DNA-methylation and expression profiling of 215 adult B-ALL patients on a single phase III trial found that MLL-rearranged B-ALL features prominent cytosine hypomethylation linked with MLL fusion protein binding, H3K79 dimethylation and transcriptional upregulation, affecting a set of MLL-fusion direct targets with oncogenic activity of which FLT3 is one E9. That is the right lineage in human patients, and it was cited in no entry of this corpus before 2026-09-21. On top of it sits a functional result: a dropout CRISPR screen over 229 HOXA9-bound peaks in MLL-r leukaemia reproducibly scores the distal enhancer of FLT3, and Cas9 editing or dCas9-KRAB silencing of that element reduces FLT3 transcription and impairs proliferation in vitro and in NSG transplants E10. And the HOXA9/MEIS1 module's mechanism at such elements is acetylation: conditional Hoxa9 activation recruits CBP/P300 and acetylates histones at H3K4me1-marked distal sites over a proto-oncogene network in which Flt3 is named E11. This graph already holds the fourth strand — that transcriptional downregulation of FLT3 is the major pathway by which low-dose pinometostat kills MLL-rearranged cells, with H3K79me2 depleted at those targets by calibrated ICeChIP-seq — in mechanism-h3k79me2-noncanonical-flt3-survival, and this amendment links to it rather than restating it.

Put together, the FLT3 locus in MLL-rearranged leukaemia is a hypomethylated, H3K79me2-marked, HOXA9-enhancer-driven, DOT1L-sensitive locus. That materially changes how this entry's central argument reads. The dossier above argues for selecting infants on FLT3 expression rather than mutation. The epigenetic layer says what that expression is made of, and it names two upstream pharmacological handles — DOT1L and the HOXA9 enhancer — that a FLT3 inhibitor does not touch.

Outbound, and this is the part that defeats the default answer. In a retroviral KMT2A::MLLT3 mouse AML model, FLT3-ITD, FLT3 N676K and NRAS G12D each remodel the chromatin accessibility landscape and the associated transcriptional networks. The mutations share a common core of chromatin changes, but each has a unique profile, with most opened peaks associating with enhancers in intronic or intergenic regions; FLT3 N676K and NRAS G12D rewire similar networks, distinct from those mediated by FLT3-ITD; motif analysis places AP-1 in the N676K/G12D leukaemias and Runx1 and Stat5a/Stat5b in the presence of FLT3-ITD; and the results were validated and extended to acute leukaemia in infancy E12. So a co-occurring FLT3 lesion is not a signalling detail layered on a fixed KMT2A-r epigenome — the epigenome is FLT3-genotype-dependent, and the two FLT3 alleles do different things to it.

What this is not. Accessibility is not a histone mark, and this amendment does not record one for FLT3. Nor does it record a mark FLT3 writes: the outbound effect is the downstream consequence of constitutive kinase signalling acting through transcription-factor networks, and the papers describe it that way. The honest summary is that FLT3 is epigenetically coupled at both ends and epigenetically active at neither.

Facts and figures
uniprotP36888
gene symbolFLT3
ensemblENSG00000122025
length AA993
alphafold IDAF-P36888-F1
cytoband13q12.2
druggabilityWell drugged. Gilteritinib (type I) and quizartinib (type II) are FDA-approved in adult AML with co-crystal structures. No FLT3-inhibitor trial in infant KMT2A-r ALL has selected patients on FLT3 expression. The pharmacodynamic half of that sentence was wrong and was corrected 2026-09-20: quizartinib was dosed to a weekly plasma-inhibitory-assay endpoint in children with relapsed AML or MLL-rearranged ALL (PMID 26920889), with complete FLT3-phosphorylation inhibition in all patients and no significant blast reduction in the FLT3 wild-type group. No ChEMBL row count is quoted here because a target-level count would not distinguish the ATP pocket from other sites.
subnuclear localizationnot applicable - FLT3 is a receptor tyrosine kinase; no nuclear or subnuclear pool is reported in any source read for this beat
shown for fusionPARTLY - the NOX/ROS compartment experiment (PMID 27870947) was performed in MV4-11, a t(4;11) KMT2A::AFF1-positive line, but the lesion imaged is FLT3-ITD rather than the fusion, and no study asks whether a KMT2A fusion alters FLT3 trafficking
epigenetic readoutNOT 'NOT APPLICABLE' - FLT3 has epigenetic content in BOTH directions, and neither of them is a mark FLT3 writes. FLT3 is a class III receptor tyrosine kinase with no chromatin domain; its substrates are non-histone, and FLT3-ITD[tiab] AND H3K9[tiab] returns 0, so no histone-mark activity is attributed to it. INBOUND, the chromatin state of the FLT3 locus: (a) in HUMAN adult MLL-rearranged B-ALL (n=215, ECOG E2993), FLT3 is one of the MLL-fusion direct targets showing prominent cytosine HYPOmethylation linked to MLL fusion protein binding, H3K79 DIMETHYLATION and transcriptional upregulation - the right lineage and the right disease class; (b) the FLT3 DISTAL ENHANCER is a HOXA9-bound element that is FUNCTIONALLY REQUIRED: a noncoding dropout CRISPR screen over 229 HOXA9-bound ChIP-seq peaks in MLL-r leukaemia scores it reproducibly, and Cas9 editing or dCas9-KRAB silencing of it reduces FLT3 transcription and impairs proliferation in vitro and in vivo; (c) Flt3 is in the proto-oncogene set that gains CBP/p300 recruitment and histone ACETYLATION on conditional Hoxa9 activation at H3K4me1-marked distal enhancers (murine, not a fusion system); (d) transcriptional downregulation of FLT3 is the MAJOR pathway by which low-dose pinometostat kills MLL-rearranged cells, with H3K79me2 markedly depleted at those targets by calibrated ICeChIP-seq - already held in mechanism-h3k79me2-noncanonical-flt3-survival. OUTBOUND, and this is the part that makes the default answer wrong: FLT3-ITD REMODELS THE CHROMATIN ACCESSIBILITY LANDSCAPE in a KMT2A::MLLT3 mouse AML model, with a mutation-SPECIFIC profile distinct from FLT3 N676K and NRAS G12D, most opened peaks at intronic/intergenic ENHANCERS, and Runx1 and Stat5a/Stat5b motifs active under ITD where AP-1 is active under N676K/G12D; the findings were validated and extended to acute leukaemia in infancy. That is co-mutation epigenetics - accessibility, not a histone mark - and it means a KMT2A-r epigenome is FLT3-genotype-dependent. CAUSE OR CONSEQUENCE: for the locus, unresolved in the usual way; for the outbound direction the perturbation is genetic and the direction is therefore established, though in KMT2A::MLLT3 and not the index fusion.
domains
Ig-like C2-type (extracel…253.0Ig-like C2-type (extracellular): 253Juxtamembrane autoinhibit…591.0Juxtamembrane autoinhibitory segment (UniProt-flagged residues): 591Protein kinase domain (sp…610.0Protein kinase domain (split): 610
most depleted leukemia lines
MOLM-13-2.181MOLM-13: -2.181SEMK2-1.715SEMK2: -1.715HB11;19-1.617HB11;19: -1.617MV4-11-1.509MV4-11: -1.509SEM-0.982SEM: -0.982MOLM-14-0.970MOLM-14: -0.97
PDB IDS
[
  "1RJB",
  "3QS7",
  "3QS9",
  "4RT7",
  "4XUF",
  "5X02",
  "6IL3",
  "6JQR",
  "7QDP",
  "7ZV9",
  "8XB1"
]
PDB verified
[
  "1RJB",
  "4XUF",
  "6JQR",
  "7ZV9"
]
domains — the numbers behind the chart
[
  {
    "name": "Ig-like C2-type (extracellular)",
    "start": 253,
    "end": 343,
    "pfam": null
  },
  {
    "name": "Juxtamembrane autoinhibitory segment (UniProt-flagged residues)",
    "start": 591,
    "end": 597,
    "pfam": null
  },
  {
    "name": "Protein kinase domain (split)",
    "start": 610,
    "end": 943,
    "pfam": null
  }
]
essentiality
{
  "depmap_common_essential": false,
  "chronos_mean": -0.145,
  "chronos_mean_leukemia_lines": -0.2789,
  "n_lines": 1243,
  "n_leukemia_lines": 57,
  "fraction_lines_depleted": 0.0088,
  "gnomad_pLI": 0.998,
  "gnomad_loeuf": 0.465,
  "gnomad_obs_lof": 46,
  "gnomad_exp_lof": 126.6,
  "mouse_ko": "Flt3-null mice are viable healthy adults with normal mature haematopoiesis but a specific deficiency in primitive B-lymphoid progenitors and impaired T-cell and myeloid reconstitution on transplantation; combined c-kit/flt3 deficiency is postnatally lethal",
  "hpa_ntpm": {
    "bone_marrow": 5,
    "lymph_node": 6.1,
    "spleen": 5.3,
    "thymus": 4.3,
    "classical_monocyte": 3.6,
    "liver": 0.4,
    "heart_muscle": 0.1,
    "testis": 0.6
  },
  "hpa_specificity": "Tissue enhanced; effectively confined to haematopoietic tissue"
}
codependency men1
{
  "r_all_lines": 0.1516,
  "r_leukemia_lines": 0.6452,
  "ci_leukemia": [
    0.3617,
    0.8102
  ],
  "p_permutation": 0.0002,
  "p_holm": 0.0052,
  "rank_in_leukemia_panel": 4,
  "n_leukemia_lines": 54
}
most depleted leukemia lines — the numbers behind the chart
[
  {
    "line": "MOLM-13",
    "fusion": "KMT2A::MLLT3",
    "flt3_itd": true,
    "chronos": -2.181
  },
  {
    "line": "SEMK2",
    "fusion": "KMT2A::AFF1",
    "flt3_itd": false,
    "chronos": -1.715
  },
  {
    "line": "HB11;19",
    "fusion": "KMT2A::MLLT1",
    "flt3_itd": false,
    "chronos": -1.617
  },
  {
    "line": "MV4-11",
    "fusion": "KMT2A::AFF1",
    "flt3_itd": true,
    "chronos": -1.509
  },
  {
    "line": "SEM",
    "fusion": "KMT2A::AFF1",
    "flt3_itd": false,
    "chronos": -0.982
  },
  {
    "line": "MOLM-14",
    "fusion": "KMT2A::MLLT3",
    "flt3_itd": null,
    "chronos": -0.97
  }
]
trial reconciliation
{
  "trial": "trial-cog-aall0631",
  "nct_note": "NCT number not verified by us in this entry; cite the publication rather than a registry id",
  "randomised_result": "no EFS difference: 36 +/- 6% (lestaurtinib, n=67) vs 39 +/- 7% (chemotherapy, n=54), p = 0.67",
  "pd_subgroup": "EFS 59 +/- 10% (FLT3 inhibition achieved) vs 28 +/- 7% (not achieved), p = 0.009",
  "evs_subgroup": "EFS 52 +/- 8% (ex vivo sensitive) vs 5 +/- 5% (resistant), p < 0.001",
  "both": "EFS 88 +/- 8% in the 17 patients both inhibited and sensitive",
  "interpretation": "compatible with a pharmacodynamic failure of a weak inhibitor rather than absence of a dependency; post-hoc and confounded, so hypothesis-generating only"
}
localization
{
  "subcellular": "wild-type FLT3 matures to the PLASMA MEMBRANE (single-pass type I). FLT3-ITD is retained intracellularly in the early secretory pathway as an immature underglycosylated species: the endoplasmic reticulum per UniProt P36888 and the compartment-targeting literature (PMID 19854140), or a Golgi-marker-positive perinuclear region per the study of endogenous receptor in patient-derived AML lines (PMID 34811450). Mutant NPM1c moves FLT3-D835Y from the surface to the ER (PMID 31186273).",
  "subnuclear": "not applicable - FLT3 is a receptor tyrosine kinase with no nuclear pool reported in any source read for this beat",
  "how_shown": "compartment-targeted constructs with global phosphoproteomics (12,186 sites); co-immunofluorescence with Golgi markers on endogenous receptor; tunicamycin/brefeldin A trafficking block in MV4-11; glycoform analysis; murine co-expression models and patient samples",
  "functional_consequence": "ER-localised FLT3-ITD activates STAT5 and PIM1/2 but not PI3K or MAPK; membrane-targeted FLT3-ITD activates MAPK and PI3K with diminished STAT5; Golgi-localised FLT3-ITD activates AKT and ERK. Quizartinib and midostaurin reduce retention and increase plasma-membrane FLT3-ITD.",
  "contradiction": "ER versus Golgi as the principal intracellular site, unresolved; resolving experiment is compartment-restricted FLT3-ITD in one endogenous-ITD background scoring STAT5, AKT and ERK together",
  "lineage_trap": "the entire literature concerns FLT3-ITD, an AML lesion; infant KMT2A-rearranged B-ALL overexpresses wild-type FLT3 or carries point mutations, and overexpressed wild-type receptor traffics to the surface, so this mechanism does not transfer",
  "node": "localization-flt3-itd-intracellular-retention-er-versus-golgi"
}
cell type
{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "expressed_in": {
    "class": 1,
    "summary": "FLT3 (CD135) marks LMPPs and early lymphoid and myeloid progenitors in normal haematopoiesis, and is expressed at high level in KMT2A-rearranged leukaemia. This is the one gene in the worklist where high expression was the ORIGINAL route to a therapeutic hypothesis, via gene-expression-based classification (PMID 12620411).",
    "note_on_evidence_class": "the expression came first and the requirement was then tested separately - which is the correct order and is worth naming, because the graph elsewhere contains expression presented as if it were requirement"
  },
  "required_in": {
    "class": 2,
    "kmt2a_rearranged_lines_mean_gene_effect": -0.962,
    "all_1258_lines_mean_gene_effect": -0.145,
    "pct_all_lines_depleted": 0.9,
    "floor_compressed": false,
    "myeloid_kmt2ar_mean": -0.859,
    "blineage_kmt2ar_mean": -1.142,
    "verdict": "the SECOND largest KMT2A-rearranged-selective dependency in the worklist after MEN1 (-0.817 against all lines), and it is LARGER in the B-lineage half than in the myeloid half (-1.142 against -0.859). That inverts the common assumption that FLT3 is a myeloid-AML target.",
    "contrast_C_is_uninterpretable_here": "MV4-11 carries an FLT3-ITD (Cellosaurus CVCL_0064, from PMID 12529668), so the same-fusion cell-type contrast for FLT3 is confounded by genotype and must not be read",
    "orthogonal_support": "FLT3 inhibition is cytotoxic in primary MLL-rearranged infant ALL (PMID 15956279), i.e. in the index cell type and in primary human material rather than in lines"
  },
  "context": "fusion context, and uniquely for this beat there is primary human infant ALL material (PMID 15956279)",
  "graph_links": [
    "cell-type-lmpp",
    "cell-type-pro-b"
  ]
}
cell state
{
  "maintains": {
    "class": 3,
    "states": [],
    "claim": "no cell-STATE claim, despite an enormous literature. The 617 records returned by the state query are overwhelmingly measurable-residual-disease monitoring of FLT3-ITD and post-transplant relapse prevention - FLT3 as a MEASUREMENT of a state, not as a maintainer of one."
  },
  "bounded_query": {
    "date": "2026-09-20",
    "tool": "NCBI esearch db=pubmed",
    "query": "(FLT3[tiab] OR \"FLT-3\"[tiab] OR \"CD135\"[tiab] OR \"FLK2\"[tiab] OR \"STK-1\"[tiab]) AND <cell-state clause> AND <leukaemia clause>",
    "count": 617,
    "controls": {
      "gene_clause_alone": 9106,
      "state_clause_alone": 96743
    },
    "reading": "a large count that is not an answer. This is the clearest case in the beat of a high recall that measures clinical monitoring rather than biology, and it is the mirror image of the bounded zeros elsewhere in this worklist."
  },
  "mechanism_of_the_expression": "MBNL1 hijacks a structured single-stranded distal DNA element to sustain FLT3 expression in KMT2A-rearranged leukaemias (PMID 42566532), which is a mechanism for WHY the expression is high in this cell type"
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.9618,
  "chronos_mean_other_leukaemia": -0.1193,
  "difference": -0.8425,
  "pan_essentiality_all_lines": -0.1466,
  "selectivity_rank": 2,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
epigenetic bounded absence
{
  "claim": "No histone-mark-writing or -erasing activity is attributed to FLT3 or to FLT3-ITD in the title/abstract index.",
  "date_measured": "2026-09-21",
  "tool": "NCBI esearch db=pubmed (eutils), querytranslation read back from each response",
  "queries": [
    {
      "query": "FLT3-ITD[tiab] AND H3K9[tiab]",
      "querytranslation": "\"FLT3-ITD\"[Title/Abstract] AND \"H3K9\"[Title/Abstract]",
      "count": 0
    },
    {
      "query": "FLT3-ITD[tiab] AND histone[tiab]",
      "querytranslation": "\"FLT3-ITD\"[Title/Abstract] AND \"histone\"[Title/Abstract]",
      "count": 58,
      "role": "positive control through a near-identical shape - the instrument returns plenty; the first twelve were title-checked and are HDAC-inhibitor combination pharmacology, an EP300-related lncRNA correspondence, and one KMT2A-r chromatin-accessibility paper (PMID 39329074, now cited here). NONE reports FLT3 or FLT3-ITD writing or erasing a histone mark."
    },
    {
      "query": "FLT3[tiab] AND H3K79[tiab]",
      "querytranslation": "\"FLT3\"[Title/Abstract] AND \"H3K79\"[Title/Abstract]",
      "count": 3,
      "role": "the inbound direction, for contrast: PMID 34263728 (the low-dose pinometostat/FLT3 pathway), PMID 23107779 (the B-ALL epigenome, now cited here) and one off-topic genetic-instability review"
    },
    {
      "query": "FLT3-ITD[tiab] AND DNMT3A[tiab] AND methylation[tiab]",
      "querytranslation": "\"FLT3-ITD\"[Title/Abstract] AND \"DNMT3A\"[Title/Abstract] AND \"methylation\"[Title/Abstract]",
      "count": 32,
      "role": "shows where the FLT3-ITD epigenetics literature actually lives: CO-MUTATION with epigenetic modifiers, not ITD-driven mark deposition"
    }
  ],
  "absence_class": "never-measured",
  "absence_class_basis": "Bounded by titles and abstracts only; supporting figures not searched (rule 80). The bound is on the ATTRIBUTION of an activity rather than on a measurement, which is the weaker kind of absence - but it is also the kind that matters here, since the risk was defaulting FLT3 to epigenetically-not-applicable and the correct entry is the opposite.",
  "why_it_matters": "This entry already records subnuclear_localization as not applicable for FLT3, correctly. The epigenetic column is NOT the same call, and recording it as not applicable would have hidden a human B-lineage methylation/H3K79me2 result at the FLT3 locus, a functionally validated HOXA9-bound FLT3 enhancer, and a mutation-specific chromatin-accessibility programme driven by FLT3-ITD in a KMT2A-rearranged model."
}
How this could bite us back
  • Resistance: RAS-pathway and FLT3 co-mutation is already documented in this graph as a route to menin-inhibitor resistance, so a menin/FLT3 combination is partly hedging against itself rather than against independent risks.
  • Essential-function toxicity: Flt3-null mice have a specific defect in primitive B-lymphoid progenitors and impaired reconstitution after transplantation. In an infant B-ALL population routinely treated with HSCT, both of those are the wrong compartments to impair.
  • Delivery and pharmacodynamics: AALL0631 shows the failure mode concretely. Most treated infants never achieved FLT3 inhibition, so a repeat trial without a measured pharmacodynamic endpoint would fail for the same reason and be read as a second refutation of the biology.
  • Selection error: the field's habit of selecting FLT3 trials on mutation status would enrol about 3% of infant KMT2A-r patients and would miss the mechanism entirely, since the lesion here is overexpression of the wild-type receptor.
  • Interpretive risk to this project: if FLT3 dependency reflects nothing more than 'KMT2A-rearranged lines need FLT3 and also need menin', then the co-dependency adds no information, and treating r = 0.65 as evidence of shared mechanism would be a category error.
  • The FLT3 locus has at least three upstream handles - DOT1L-deposited H3K79me2, a HOXA9-bound distal enhancer, and cytosine hypomethylation - so transcriptional escape is the resistance mechanism to expect from a FLT3 kinase inhibitor in this disease, distinct from the kinase-domain mutations that drive resistance in FLT3-mutant AML. An expression-selected trial should measure FLT3 transcript on therapy, not only plasma inhibitory activity.
  • Conversely, a demethylating agent given to raise or lower FLT3 in KMT2A-r disease would be acting on a locus whose methylation direction is subtype-split: fusion-target hypomethylation in adult MLLr against extensive promoter hypermethylation in infant t(4;11). The same drug could move the locus in opposite directions in the two populations.
Still open
  1. Would gilteritinib or quizartinib, dosed to a measured plasma FLT3-inhibition endpoint and given to infants selected on FLT3 expression rather than mutation, reproduce the 88% EFS seen in AALL0631's inhibited-and-sensitive subgroup? This is the single highest-value untested question in this dossier. NARROWED 2026-09-20: the pharmacodynamic-endpoint half of this question was answered in relapsed children including MLL-rearranged ALL (PMID 26920889) and the answer was unfavourable for FLT3 wild-type patients. What remains open is specifically expression-selected, infant, frontline.
  2. Are SEM, SEMK2 and HB11;19 genuinely FLT3 wild-type? Cellosaurus records no FLT3 variant, but absence of a record is not sequencing. Confirming this would turn a suggestive observation into the central argument for expression-based selection.
  3. Is the FLT3-MEN1 co-dependency mechanistic or coincidental? Both are required by KMT2A-rearranged lines, so the correlation may carry no information beyond that. Testing whether menin inhibition changes FLT3 dependency, or vice versa, in an isogenic pair would distinguish them.
  4. Does menin inhibition upregulate FLT3 signalling as an escape route? RAS-pathway and FLT3 co-mutation is already recorded in this graph as a menin-inhibitor resistance mechanism, which makes FLT3 a candidate combination partner and a candidate resistance node at the same time.
  5. Is ex vivo FLT3-inhibitor sensitivity prognostic independently of treatment? If archived AALL0631 chemotherapy-arm samples were assayed, the 5% EFS figure could be tested as a biomarker rather than a drug effect, which would settle the confounding directly.
  6. Does a co-occurring FLT3 lesion change the KMT2A-r epigenome in the INDEX fusion, and in B-lineage? The accessibility-remodelling result is murine, retroviral and KMT2A::MLLT3, yet activating FLT3 mutation is rare in infant KMT2A-r disease with no ITD observed in AALL0631 - so the most important version of this question is whether high wild-type FLT3 EXPRESSION, which is near-universal there, does anything to chromatin at all.
  7. Is the HOXA9-bound FLT3 distal enhancer present and required in a human KMT2A::AFF1 cell? The screen that validated it did not resolve the fusion partner, and this graph already carries exp-noncoding-dependency-crispr-screen-kmt2a-aff1 as the index-fusion version. Combining that screen with a DOT1L inhibitor arm would also test whether the enhancer is one of the H3K79-dependent KEE class.
  8. Would a DOT1L inhibitor and a FLT3 inhibitor be redundant rather than additive? Low-dose pinometostat kills MLL-rearranged cells largely BY downregulating FLT3, so the two agents may occupy the same node. An epistasis experiment would say whether the combination the DOT1L literature proposes is a combination at all.
Every source behind this entry (13)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    FLT3 is not DepMap common-essential and is depleted in only 0.9% of 1,243 lines (mean Chronos -0.145; -0.279 across 57 leukemia lines), yet the six most FLT3-depleted leukemia lines are MOLM-13, SEMK2, HB11;19, MV4-11, SEM and MOLM-14; gnomAD reports pLI 0.998 and LOEUF 0.465; HPA reports tissue-enhanced haematopoietic expression with bone marrow 5.0 nTPM against heart muscle 0.1.

    What this evidence cannot show

    Per-line Chronos values retrieved through Open Targets and means computed by us; DepMap release version not stated by the API and unverified. FLT3 has data in 1,243 rather than 1,258 lines, so the panel differs slightly from other genes here. A gene expressed in few lines gives an unstable dependency distribution regardless of biology, and FLT3 is expressed in few lines by design.

    Evidence that argues against this (1)
    • A Chronos mean mixed across a KMT2A-rearranged subset and a comparator arm must be lineage-split before it is stated: the index arm is myeloid-enriched, so a difference between the two mixes fusion status with lineage. Ten or eleven index lines and four KMT2A::AFF1 lines also put wide intervals on every figure quoted. Depletion in 0.9% of lines is the more important number: FLT3 is a narrow dependency, and this project has since measured that the scan's statistic rewards narrowness — on raw effect size FLT3 ranks second behind MEN1 rather than first. Oncofusion knowledge graph (internal analysis) 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    COG AALL0631 found no EFS difference between chemotherapy plus lestaurtinib (n = 67, 3-year EFS 36 +/- 6%) and chemotherapy alone (n = 54, 39 +/- 7%, p = 0.67) in newly diagnosed KMT2A-rearranged infant ALL, but within the lestaurtinib arm EFS was 59 +/- 10% for patients achieving plasma FLT3 inhibition versus 28 +/- 7% for those not (p = 0.009), 52 +/- 8% for ex vivo sensitive versus 5 +/- 5% for resistant (p < 0.001), and 88 +/- 8% for the 17 patients who were both.

    PMID 33623141DOI
    What this evidence cannot show

    The headline randomised comparison is negative. The pharmacodynamic and ex-vivo-sensitivity comparisons are post-hoc, non-randomised subgroup analyses within one arm, with no correction for multiplicity reported. Achieving plasma inhibition covaries with organ function, nutrition and compliance, all independently prognostic in infants. Ex vivo sensitivity is plausibly a prognostic biomarker of disease biology rather than a marker of drug benefit; the 5 +/- 5% EFS in the resistant group is too extreme to be produced by drug assignment alone. These subgroups generate a hypothesis; they do not test one.

    Evidence that argues against this (1)
    • The primary randomised result of this same trial is the strongest published evidence against FLT3 inhibition as a strategy in infant KMT2A-r ALL, and it is the result with the least confounding. An erratum was published for this paper. Brown PA et al. 2021
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PKC412 is differentially cytotoxic to leukemia cells with MLL translocations in which FLT3 is activated either by overexpression of the wild-type receptor or by mutation, and activation-loop FLT3 mutations were found in only 5 of 30 MLL-rearranged cases.

    Inhibition of FLT3 in MLL. Validation of a therapeutic target identified by gene expression based classification. · Armstrong SA, Kung AL, Mabon ME, Silverman LB, Stam RW, Den Boer ML, Pieters R, Kersey JH, Sallan SE, Fletcher JA, Golub TR, Griffin JD, Korsmeyer SJ · Cancer Cell 2003
    PMID 12620411DOI
    What this evidence cannot show

    PKC412 (midostaurin) is a multikinase inhibitor, so cytotoxicity in these assays is not attributable to FLT3 blockade alone; this graph's own co-mutation entry records that at concentrations fully suppressing FLT3 phosphorylation, cytotoxicity remained fractional. The 5-of-30 mutation figure comes from a cohort not restricted by age and is widely over-quoted for infants, where activation-loop mutation is about 3% and ITD has not been observed.

    Evidence that argues against this (1)
    • PKC412 differential cytotoxicity is an ex vivo and cell-line result from 2003, and the clinical test in this exact population was run and was negative: lestaurtinib added to chemotherapy in newly diagnosed KMT2A-rearranged infant ALL did not improve outcome. Leukemia 2021
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    MOLM-13 and MV4-11 carry FLT3 internal tandem duplications, whereas SEM, SEMK2 and HB11;19 have KMT2A fusions recorded with no FLT3 variant of any kind listed.

    What this evidence cannot show

    Absence of a recorded FLT3 variant in Cellosaurus is absence of a curated record, not proof of wild-type status; Cellosaurus curates from published reports and coverage is uneven across lines. The claim that these lines are FLT3 wild-type should be verified by sequencing before it is used to support a clinical argument. Note also that the FLT3-ITD lines MOLM-13 and MV4-11 are the two most depleted, so ITD status clearly amplifies the dependency even if it is not necessary for it.

    Evidence that argues against this (1)
    • Cellosaurus recording no FLT3 variant is an absence in a curation, not a sequenced negative: variant records depend on what was deposited for that line. The argument the entry builds on it — that FLT3 dependency is ITD-independent — is supported better by the per-line effect sizes than by the annotation. Oncofusion knowledge graph (internal analysis) 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Across 54 DepMap leukemia lines FLT3 correlates with MEN1 at r = 0.645 (95% CI 0.362-0.810, permutation p = 0.0002), surviving Holm correction and ranking fourth behind only KMT2A, PSIP1 and DOT1L, while across the full 1,208-line panel the correlation is only r = 0.15.

    What this evidence cannot show

    Our own computation, not peer reviewed, in 54 leukemia lines with a wide bootstrap interval. FLT3 and MEN1 are both required by KMT2A-rearranged lines, so their correlation may reflect nothing more than both being needed by the same handful of lines rather than any shared process. No lineage adjustment within the leukemia subset.

    Evidence that argues against this (1)
    • A selectivity difference expressed as a rank inherits this project's own correction to its instrument: the scan's statistic divides by comparator standard deviation, which ranks NARROWNESS rather than effect size at Spearman −0.92 in the depleted tail. A rank from that scan is not an effect-size ordering, and 167 entries were written before that was measured. A co-dependency of 0.645 with MEN1 across 54 leukaemia lines is also consistent with both genes tracking the same lineage rather than a shared mechanism. Oncofusion knowledge graph (internal analysis) 2026
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Mice deficient in flk2/Flt3 develop into healthy adults with normal mature haematopoietic populations but specific deficiencies in primitive B-lymphoid progenitors, and their stem cells show reduced T-cell and myeloid reconstitution after transplantation.

    Targeted disruption of the flk2/flt3 gene leads to deficiencies in primitive hematopoietic progenitors. · Mackarehtschian K, Hardin JD, Moore KA, Boast S, Goff SP, Lemischka IR · Immunity 1995
    PMID 7621074DOI
    What this evidence cannot show

    Germline knockout in mouse, which allows developmental compensation that acute pharmacological inhibition does not. The B-lymphoid progenitor deficit is directly relevant to an infant B-ALL population in which B-cell recovery matters, and the transplantation deficit matters in a disease treated with HSCT.

    Evidence that argues against this (1)
    • This is a mouse knockout describing what the gene does in normal development or normal haematopoiesis, over the animal's whole life, with the protein entirely absent. A drug gives partial reversible inhibition in an adult for weeks. The experiment bounds the biological requirement and not the therapeutic window. Normal adult mice with deficits confined to primitive B-lymphoid progenitors is the phenotype most relevant to an infant B-ALL patient, and it is the one the entry files last. Leukemia 2021
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Co-crystal structures of the FLT3 kinase domain with the approved inhibitors gilteritinib (6JQR, 2.2 A) and quizartinib (4XUF, 3.2 A) exist, alongside the apo receptor (1RJB, 2.1 A) and a complex with a monomeric FLT3-ligand variant (7ZV9, 4.5 A).

    What this evidence cannot show

    We verified accession existence, title, method and resolution only; we did not re-refine the models or independently confirm ligand occupancy. 7ZV9 at 4.5 A is low resolution for interpreting side-chain contacts.

    Evidence that argues against this (1)
    • Counting PDB entries measures how tractable a protein was to CRYSTALLISE, which correlates with being small, ordered and well-behaved rather than with being druggable. What decides design value is whether the deposited density covers the site that matters, at a resolution that supports modelling, in the relevant state — and this graph has repeatedly found structures described in the direction that overstates what is resolved. Two approved inhibitors co-crystallised means FLT3's structural and chemical problems are solved; what failed in this disease was the clinical result, which no structure addresses. Leukemia 2021
  8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt P36888 annotates human FLT3 as 993 residues with an Ig-like C2-type domain at 253-343, a protein kinase domain at 610-943, and residues 591-597 in the juxtamembrane region as important for maintaining the kinase in an inactive state in the absence of bound ligand.

    What this evidence cannot show

    UniProt feature boundaries are curated annotations, in places predicted rather than experimentally mapped, and different sources place the juxtamembrane domain over a wider range than the 591-597 UniProt flags. Exon-level ITD boundaries should be taken from a genomic source, not from these protein coordinates.

    Evidence that argues against this (1)
    • A UniProt domain table is a curated annotation. The juxtamembrane region 'important' at 591-597 is the ITD hotspot, and its functional description in UniProt is inference from the literature rather than an independent measurement. UniProt Consortium 2026
  9. E-quizartinib-taclestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.journal-article

    A phase I study dosed quizartinib, a potent and selective FLT3 inhibitor, on top of high-dose cytarabine and etoposide in children with relapsed AML or MLL-rearranged ALL, using weekly plasma inhibitory assay testing as a pharmacodynamic endpoint. FLT3 phosphorylation was completely inhibited in all patients, one of 6 had a dose-limiting toxicity at 40 mg/m2/day and one of 9 at the highest tested 60 mg/m2/day, and of 17 response-evaluable patients there were 2 complete responses, 1 CRp, 1 CRi, 10 stable disease and 3 progressive disease. FLT3-ITD patients, but not FLT3 wild-type patients, had significantly lower blast counts after quizartinib.

    A Phase I Study of Quizartinib Combined with Chemotherapy in Relapsed Childhood Leukemia: A Therapeutic Advances in Childhood Leukemia & Lymphoma (TACL) Study · Cooper TM, Cassar J, Eckroth E, Malvar J, Sposto R, Gaynon P, Chang BH, Gore L, August K, Pollard JA, DuBois SG, Silverman LB, Oesterheld J, Gammon G, Magoon D, Annesley C, Brown PA · Clinical Cancer Research 2016
    PMID 26920889DOI
    What this evidence cannot show

    Read from the structured abstract on 2026-09-20; **the full text was not reachable** - Europe PMC returns no PMCID and `elink` offers only `pubmed_pmc_refs`, which is the reverse-citation list and not the article, a trap already logged in this project. So the **ages of the enrolled children and the number who had MLL-rearranged ALL are unknown here**, and 'children' must not be read as 'infants'. Established as a fact about what this trial reported, under the grading rule; it is a phase I dose-finding study of 17 response-evaluable patients on combination chemotherapy, so no efficacy conclusion of any kind transfers, in either direction.

    Evidence that argues against this (1)
    • This entry asserted for several cycles that no modern selective FLT3 inhibitor had been dosed to a pharmacodynamic endpoint in this population, which was wrong by ten years. It was found by diffing PubMed hits against the PMIDs each entry already cites - the same instrument returned hits for three other absence claims and all three turned out to be papers the entry already carried, so the filter is what made this one visible. Oncofusion knowledge graph (internal analysis) 2026
  10. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In HUMAN adult MLL-rearranged B-ALL, integrative DNA-methylation and gene-expression profiling of 215 patients enrolled on the ECOG E2993 phase III trial plus normal control B cells found prominent cytosine hypomethylation linked with MLL fusion protein binding, H3K79 dimethylation and transcriptional upregulation, affecting a set of known and newly identified MLL fusion direct targets with oncogenic activity of which FLT3 and BCL6 are named; BCL6 blockade or loss of function suppressed proliferation and survival of MLL-rearranged leukaemia cells.

    PMID 23107779DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. ADULT B-ALL, not infant, and 'MLLr' is a partner-mixed class - the abstract does not state how many of the MLLr cases are KMT2A::AFF1, so this is B-lineage human evidence for the disease class and NOT index-fusion-specific (rule 86). The hypomethylation, fusion binding and H3K79me2 association is a correlative integration of three genome-wide layers in patient material; nothing in the abstract perturbs methylation or H3K79me2 and then measures FLT3, so the arrow into FLT3 is association. The functional arm of the paper is BCL6, not FLT3. NOT CITED anywhere in this corpus before 2026-09-21, verified with scripts/check-citation-reuse.py.

    Evidence that argues against this (1)
    • The methylation direction is subtype-dependent and the index fusion sits on the opposite side of it, so 'FLT3 is hypomethylated in MLLr' must not be generalised. Promoter CpG island methylation separates MLL-rearranged INFANT ALL into subtypes in which t(4;11) and t(11;19) are extensively HYPERmethylated while t(9;11) and MLL-germline infant ALL resemble normal bone marrow. Global hypomethylation of fusion targets in an adult MLLr cohort and extensive promoter hypermethylation in infant t(4;11) are not contradictory - one is target-locus, the other promoter-CpG-island, and the cohorts differ in age - but an entry that recorded only the first would give an infant-ALL reader the wrong picture of this disease's methylome. Stumpel DJ et al. 2009
  11. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The FLT3 DISTAL ENHANCER is a functionally required HOXA9-bound regulatory element in MLL-r leukaemia: dropout CRISPR screens against 229 HOXA9-bound peaks identified by ChIP-seq return reproducible noncoding hits including one in the distal enhancer of FLT3 and one in an intron of CDK6, and Cas9 editing or dCas9-KRAB silencing of the HOXA9-bound sites significantly reduces the corresponding gene transcription and impairs cell proliferation in vitro and in vivo on transplantation into NSG mice, with accompanying chromatin-accessibility and chromatin-conformation changes.

    PMID 38016946DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. 'MLL-r' is not resolved to a partner in the abstract, so this is not established for KMT2A::AFF1; this graph's exp-noncoding-dependency-crispr-screen-kmt2a-aff1 exists precisely because the index-fusion version has not been done. A dCas9-KRAB result at an enhancer shows the element is required for transcription, not that HOXA9 binding is what makes it work - the screen was designed around HOXA9-bound peaks, so HOXA9 dependence is an assumption of the design rather than an output of it. No histone mark is measured at the FLT3 enhancer in the abstract; 'enhancer' here is a HOXA9 ChIP-seq peak plus accessibility, not an H3K4me1/H3K27ac call.

    Evidence that argues against this (1)
    • The enhancer's requirement is established; the inference that removing HOXA9 would therefore close it is not, and this entry's own companion gene contradicts that step. Mll-AF9 knock-in mice develop myeloid leukaemia with the same incidence and time course with or without Hoxa9, differing only in phenotype, because Hoxa5-Hoxa10 are co-upregulated and functionally overlapping. A HOXA9-bound enhancer in a redundant Hox code may be occupied and driven by a paralogue when HOXA9 is removed, so the element's requirement does not transfer to HOXA9 as a target. Kumar AR et al. 2004
  12. E11weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    Flt3 is named in the proto-oncogene network activated when Hoxa9 is conditionally activated, a step accompanied by CBP/P300 recruitment and histone acetylation: Hoxa9 and Meis1 co-bind hundreds of highly conserved, mostly TSS-distal sites carrying high H3K4 monomethylation and CBP/P300 binding characteristic of enhancers, and conditional Hoxa9 activation is associated with CBP/P300 recruitment, histone acetylation and transcriptional activation of Erg, Flt3, Lmo2, Myb and Sox4.

    Identification and characterization of Hoxa9 binding sites in hematopoietic cells. · Huang Y, Sitwala K, Bronstein J, et al. · Blood 2012
    PMID 22072553DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. The weakest of the four inbound strands and marked weak for three reasons: MURINE haematopoietic cells with conditional Hoxa9 and NOT a KMT2A-fusion system; the acetylation is reported for the activated network as a class, so it is not established that Flt3's own enhancer is the acetylated element; and the acetylated residue is not specified in the abstract, so no specific acetyl mark can be asserted at the FLT3 locus. It is recorded because it names the enzyme - CBP/p300 - that the HOXA9 module works through, which is the only mechanistic bridge available between the HOXA9-bound FLT3 enhancer and any mark.

    Evidence that argues against this (1)
    • A rival and better-controlled reading of accessibility change at the FLT3 locus in this disease does not involve HOXA9 at all: in a KMT2A::MLLT3 model the chromatin opened at intronic and intergenic enhancers is driven mutation-specifically by activating FLT3 and NRAS alleles, with Runx1/Stat5 motifs under FLT3-ITD and AP-1 under FLT3 N676K and NRAS G12D. So enhancer opening around signalling genes in a KMT2A-rearranged cell has at least two candidate drivers, and a Hoxa9-activation experiment in normal murine progenitors cannot distinguish them. Zhang Q et al. 2024
  13. E12strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    FLT3-ITD REMODELS CHROMATIN IN A KMT2A-REARRANGED CONTEXT, mutation-specifically: in a retroviral KMT2A::MLLT3 mouse AML model, FLT3-ITD, FLT3 N676K and NRAS G12D each remodelled the chromatin accessibility landscape and associated transcriptional networks, sharing a common core of chromatin changes while each exhibiting unique profiles with most opened peaks associating with enhancers in intronic or intergenic regions; FLT3 N676K and NRAS G12D rewired similar networks distinct from FLT3-ITD, motif analysis implicated AP-1 under FLT3 N676K and NRAS G12D and Runx1 and Stat5a/Stat5b under FLT3-ITD, and the results were validated and extended to acute leukaemia in infancy.

    PMID 39329074DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. MURINE and RETROVIRAL, driven by KMT2A::MLLT3 rather than the index fusion, and retroviral overexpression sets fusion and mutant allele levels artificially - rule 73 applies. The readout is chromatin ACCESSIBILITY (ATAC-type) plus motif enrichment, NOT a histone mark: no H3K27ac, H3K4me1 or H3K79me2 measurement is reported in the abstract, so 'FLT3-ITD changes enhancer marks' would overstate it. Motif enrichment names candidate transcription factors, not occupancy. The infant validation is stated in the abstract without its extent given. NOT CITED anywhere in this corpus before 2026-09-21.

    Evidence that argues against this (1)
    • The chromatin-remodelling result invites the reading that FLT3 lesions define a distinct, targetable epigenetic subgroup, and the one randomised test in the right population does not support acting on FLT3 genotype. COG AALL0631 randomised 121 KMT2A-rearranged infants and found no event-free-survival difference with lestaurtinib (36 +/- 6% versus 39 +/- 7%, p = 0.67) - and, decisively for this point, activating FLT3 mutation is rare in infant KMT2A-r disease with no ITD observed in that cohort, so the ITD-specific chromatin programme described in a mouse model may have little counterpart in the patients this project is about. The trial's pharmacodynamic subgroups rescue FLT3 as an expression-selected target, not as a genotype-selected one. Brown PA et al. 2021
Connected to